[
https://issues.apache.org/jira/browse/HBASE-18300?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17567487#comment-17567487
]
Bryan Beaudreault commented on HBASE-18300:
-------------------------------------------
[~ram_krish] did you end up doing anything here? I'm thinking about this. I
think it would be helpful with S3 backed clusters or other cloud-based
clusters. Let's say you have huge disk requirements but low CPU requirements
(low req/s). In cloud, you may not have access to perfect server configuration
and options will be limited to cut cost. Often on large clusters like this, you
may have lots of old data which is not frequently accessed. TieredBucketCache
could help here. I'm imagining something like:
* L1: 4-5gb small to keep GC costs down
* L2: off-heap BucketCache with 50-150gb
* L3: 500gb or >1TB of direct attached NVME SSDs (in AWS this might be i4i or
other similar)
* HDFS: backed by S3 or slow EBS drives
Today we can accomplish this with BucketCache, but we miss out on the big L2
off-heap which could really help accelerate.
An alternative configuration would use fast EBS drives for L3, and slow
spinning disks for HDFS (in aws, like d3 instance type).
> Implement a Multi TieredBucketCache
> -----------------------------------
>
> Key: HBASE-18300
> URL: https://issues.apache.org/jira/browse/HBASE-18300
> Project: HBase
> Issue Type: New Feature
> Components: BucketCache
> Affects Versions: 2.0.0
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Priority: Major
>
> We did an internal brainstorming to study the feasibility of this. Some of
> our recent tests on SSDs like Optane shows that they are vastly faster in
> randomreads and can act as effective caches.
> In the current state we have a single tier of Bucket cache and the bucket
> cache can either be offheap or configured to work with file mode. (The file
> mode can have multiple files backing it).
> So this model restricts us from using either the memory or the file and not
> both.
> With the advent of faster devices like Optane SSDs, NVMe based devices it is
> better we try to utilize all those devices and try using them for the bucket
> cache so that we can avoid the impact of slower devices where the actual data
> resides on the HDFS data nodes.
> Combined with this we can allow the user to configure the caching layer per
> family/table so that one can effectively make use of the caching tiers.
> Can upload a design doc here. Before that, would like to know the suggestions
> here. Thoughts!!!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
